Hello,
I think I've got it:
The input charset in latin1, the po charset is UTF-8.
Po.pm::gettext tries to find a latin1 string in the PO, and fails. It then
directly prints the latin1 string in the output document (probably with an
additional UTF-8 to UTF-8 re-encoding).
Here is a patch. A review is welcome.
I don't commit it right known, Alioth has some troubles.
Regards,
--
Nekral